CN103327346B - Stereo matching device for judging concave and protruding blocks and method thereof - Google Patents
Stereo matching device for judging concave and protruding blocks and method thereof Download PDFInfo
- Publication number
- CN103327346B CN103327346B CN201210077582.4A CN201210077582A CN103327346B CN 103327346 B CN103327346 B CN 103327346B CN 201210077582 A CN201210077582 A CN 201210077582A CN 103327346 B CN103327346 B CN 103327346B
- Authority
- CN
- China
- Prior art keywords
- reliability
- block
- numerical value
- distribution map
- parallax
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
A stereo matching device is applied to a stereo display system to judge concave and protruding blocks. The stereo matching device comprises a receiving module which receives first visual angle frames and second visual angle frames, a calculating module, a feature capturing module and an estimating module. The calculating module produces a parallax distribution diagram which contains a plurality of parallax numerical values. The feature capturing module generates a plurality of feature distribution diagrams for all the blocks, and each feature distribution diagram includes a plurality of feature numerical values corresponding to each block. The estimating module comprises a reliability degree calculating unit which calculates the feature reliability degrees of all the blocks according to the feature distribution diagrams and a comparison unit which compares the feature reliability degree of each block with a reliability degree critical value to filter a plurality of unqualified blocks so that a candidate block can be further produced to judge the concave and protruding blocks.
Description
Technical field
Content of the present invention relates to a kind of stereo display technique, and relates to a kind of Stereo matching devices and methods therefor especially.
Background technology
The Stereo Matching Technology of image is by the one widely adopted in the technology producing 3-dimensional image data.The method of Stereo matching is in two images of different visual angles shooting, find the pixel of match (meaning and same position), and utilize the parallax between match point, principle according to trigonometric function calculates the degree of depth and the shape of object in image, and then the image of reconstruction of three-dimensional.
When applying Stereo Matching Technology, two important parameters need be estimated, meaning and recessed block and convex region block.The present technology of part adopts universe search method such as figure segmentation (graph-cut) algorithm or belief propagation (belief propagation) to carry out Stereo matching.But, no matter these algorithms have quite high cost on computing time or hardware resource complexity.Further, the result that these algorithms calculate, its reliability also cannot reach the level that can trust.
Therefore, how designing a new Stereo matching devices and methods therefor to overcome above-mentioned shortcoming, is an industry problem demanding prompt solution for this reason.
Summary of the invention
Therefore, the one side of content of the present invention is to provide a kind of Stereo matching (stereo matching) device, be applied to judge recessed (concave) block and convex (convex) block in solid (stereoscopic) display system, wherein Stereo matching device at least comprises: receiver module, computing module, feature extraction module and estimating module.Receiver module receives the first visual angle frame and the second visual angle frame and splits the first visual angle frame is multiple block.Computing module is in order to produce parallax distribution map (disparity map), and wherein parallax distribution map comprises multiple parallax numerical value, respectively corresponding block one of them the first visual angle frame and the parallax amount of the second visual angle interframe.Feature extraction module is to each onblock executing feature extraction program, and to produce multiple characteristic profile, wherein characteristic profile respectively comprises multiple character numerical value, with each block of correspondence.Estimating module comprises: reliability calculating unit and comparing unit.Reliability calculating unit calculates the feature reliability of each block according to characteristic profile, and wherein feature reliability is the summation of the character numerical value of each characteristic profile.Comparing unit compares with the multiple defective block of filtering according to the feature reliability of each block and at least one reliability critical value, the multiple candidate block of further generation, and according to there is in candidate block maximum disparity numerical value and minimum parallax numerical value person judges recessed block and convex region block respectively.
According to content one embodiment of the present invention, wherein feature extraction module comprises: chroma is as the criterion color segmentation (hue-based color segment) unit and edge detection unit.The chroma color segmentation unit that is as the criterion produces color segmentation distribution map, comprises multiple color character numerical value.Edge detection unit produces edge distribution figure, comprises multiple edge feature numerical value.
According to another embodiment of content of the present invention, wherein feature extraction module also comprises position analysis unit, in order to produce attentiveness distribution map, comprises multiple attentiveness character numerical value.
According to the another embodiment of content of the present invention, wherein reliability calculating unit comprises: the similar reliability calculating unit of parallax, edge reliability calculating unit and coupling cost (matching-cost) reliability calculating unit.Parallax similar reliability calculating unit produces similar reliability distribution map according to color segmentation distribution map and parallax distribution map, and wherein similar reliability distribution map comprises multiple similar reliability numerical value, respectively correspond to block one of them.Edge reliability calculating unit produces edge reliability distribution map according to edge distribution figure, and wherein reliability distribution map in edge comprises multiple edges reliability numerical value, respectively correspond to block one of them.Coupling cost reliability calculating unit produces coupling cost reliability distribution map according to the least absolute value sum of the deviations (minimalsum of absolute difference) of each block, wherein mate cost reliability distribution map and comprise multiple coupling cost reliability numerical value, respectively correspond to one of them coupling cost intensity of block.Wherein feature reliability is the sum total of similar reliability numerical value, edge reliability numerical value and the coupling cost reliability numerical value that block is corresponding.Wherein reliability calculating unit comprises attentiveness reliability calculating unit, in order to produce attentiveness reliability distribution map according to attentiveness distribution map, wherein attentiveness reliability distribution map comprises multiple attentiveness reliability numerical value, respectively correspond to one of them attentiveness intensity of block, feature reliability is the sum total of similar reliability numerical value, edge reliability numerical value, coupling cost reliability numerical value and the attentiveness reliability numerical value that block is corresponding.
According to a content of the present invention embodiment again, wherein computing module calculates least absolute value sum of the deviations to produce parallax distribution map according to each multiple corresponding block in each block of the first visual angle frame and the second visual angle frame.
According to the embodiment that content of the present invention has more, its concave block corresponds to the maximum disparity numerical value in parallax distribution map, and convex region block corresponds to the minimum parallax numerical value in parallax distribution map.
According to the embodiment that content of the present invention has again, wherein receiver module also comprises low pass filter and reduces sampler (down sampler), to perform low-pass filtering program to the first visual angle frame and the second visual angle frame and to reduce sample program respectively.
The another aspect of content of the present invention is to provide a kind of solid matching method, be applied to the Stereo matching device in three-dimensional display system, to judge recessed block and convex region block, wherein solid matching method at least comprises: receive the first visual angle frame and the second visual angle frame; Splitting the first visual angle frame is multiple block; Produce parallax distribution map, wherein parallax distribution map comprises multiple parallax numerical value, respectively corresponding block one of them the first visual angle frame and a parallax amount of the second visual angle interframe; Perform feature extraction program, to produce multiple characteristic profile, wherein characteristic profile respectively comprises multiple character numerical value, with each block of correspondence; Calculate the feature reliability of each block according to characteristic profile, wherein feature reliability is the summation of the character numerical value of each characteristic profile.Compare with the multiple defective block of filtering according to the feature reliability of each block and at least one reliability critical value, produce multiple candidate block further; And according to there is in candidate block maximum disparity numerical value and minimum parallax numerical value person judges recessed block and convex region block respectively.
According to content one embodiment of the present invention, the step wherein performing feature extraction program also comprises: produce color segmentation distribution map, comprise multiple color character numerical value; And produce edge distribution figure, comprise multiple edge feature numerical value.
According to another embodiment of content of the present invention, the step wherein performing feature extraction program also comprises: produce attentiveness distribution map, comprise multiple attentiveness character numerical value.
According to the another embodiment of content of the present invention, wherein the step of calculated characteristics reliability also comprises: produce similar reliability distribution map according to color segmentation distribution map and parallax distribution map, wherein similar reliability distribution map comprises multiple similar reliability numerical value, respectively correspond to block one of them; Produce edge reliability distribution map according to edge distribution figure, wherein reliability distribution map in edge comprises multiple edges reliability numerical value, respectively correspond to block one of them; Produce coupling cost reliability distribution map according to the least absolute value sum of the deviations of each block, wherein mate cost reliability distribution map and comprise multiple coupling cost reliability numerical value, respectively correspond to one of them coupling cost intensity of block; And similar reliability numerical value, edge reliability numerical value and the coupling cost reliability numerical value corresponding to block adds up to produce feature reliability.The step of calculated characteristics reliability also comprises: produce attentiveness reliability distribution map according to attentiveness distribution map, wherein attentiveness reliability distribution map comprises multiple attentiveness reliability numerical value, respectively correspond to one of them attentiveness intensity of block, feature reliability is the sum total of similar reliability numerical value, edge reliability numerical value, coupling cost reliability numerical value and the attentiveness reliability numerical value that block is corresponding.
According to a content of the present invention embodiment again, the step wherein producing parallax distribution map also comprises: calculate least absolute value sum of the deviations according to each multiple corresponding block in each block of the first visual angle frame and the second visual angle frame.
According to the embodiment that content of the present invention has more, its concave block corresponds to the maximum disparity numerical value in parallax distribution map, and convex region block corresponds to the minimum parallax numerical value in parallax distribution map.
According to the embodiment that content of the present invention has again, the step wherein receiving the first visual angle frame and the second visual angle frame also comprises respectively to the first visual angle frame and the second visual angle frame execution low-pass filtering program and minimizing sample program.
The advantage applying content of the present invention be by, and reach above-mentioned object easily.
Accompanying drawing explanation
For the above-mentioned of content of the present invention and other object, feature, advantage and embodiment can be become apparent, being described as follows of institute's accompanying drawings:
Fig. 1 is in content one embodiment of the present invention, a kind of calcspar of Stereo matching device;
Fig. 2 is in content one embodiment of the present invention, the more detailed calcspar of receiver module;
Fig. 3 is in content one embodiment of the present invention, the more detailed calcspar of computing module;
Fig. 4 is in content one embodiment of the present invention, the more detailed calcspar of feature extraction module;
Fig. 5 is in content one embodiment of the present invention, the more detailed calcspar of estimating module; And
Fig. 6 is in content one embodiment of the present invention, a kind of flow chart of solid matching method.
[main element label declaration]
1: Stereo matching device 10: receiver module
11: the first visual angle frames 12: computing module
13: the second visual angle frames 14: feature extraction module
16: estimating module 20: low pass filter
22: reduce sampler 30: coupling cost computing unit
32: disparity computation unit 34: parallax precision unit
40: chroma is as the criterion color segmentation unit 42: edge detection unit
44: position analysis unit 50: reliability calculating unit
500: the similar reliability calculating unit 502 of parallax: edge reliability calculating unit
506: coupling cost reliability calculating unit 504: attentiveness reliability calculating unit
600: solid matching method 508: reliability summation unit
52: comparing unit
601-607: step
Embodiment
Please refer to Fig. 1.Fig. 1 is in content one embodiment of the present invention, a kind of calcspar of Stereo matching device 1.Stereo matching device 1 in a three-dimensional display system (not illustrating), to judge recessed block and convex region block.Wherein Stereo matching device 1 comprises: receiver module 10, computing module 12, feature extraction module 14 and estimating module 16.
Fig. 2 is in content one embodiment of the present invention, the more detailed calcspar of receiver module 10.Receiver module 10 receives the first visual angle frame 11 and the second visual angle frame 13.In an embodiment, the first visual angle frame 11 and the second visual angle frame 13 wherein one be LOOK LEFT frame another one then for LOOK RIGHT frame, wherein LOOK LEFT frame is with thinking that the left eye of observer received, and LOOK RIGHT frame is with thinking that the right eye of observer received.In the present embodiment, receiver module 10 comprises low pass filter 20 further and reduces sampler 22, to perform low-pass filtering program to the first visual angle frame 11 and the second visual angle frame 13 respectively and to reduce sample program.Receiver module 10 further segmentation the first visual angle frame 11 (or second visual angle frame 13) is multiple block.
In an embodiment, the block of these segmentations respectively has identical size, and respectively comprises several pixel.For example, the first visual angle frame 11 can be receiver module 10 and is divided into several block respectively with 5x5 pixel.And in other embodiment, the first visual angle frame 11 can be the block that receiver module 10 is divided into other size.In other embodiment, the first visual angle frame 11 according to the color of the first visual angle frame 11 or marginal information, can be divided into several object by image patterning method by receiver module 10.
Computing module 12 is in order to produce parallax distribution map.Wherein parallax distribution map comprises multiple parallax numerical value, respectively corresponding block (or object) one of them the first visual angle frame 11 and the second visual angle frame 13 between parallax amount.Fig. 3 is in content one embodiment of the present invention, the more detailed calcspar of computing module 12.Computing module 12 comprises coupling cost computing unit 30, disparity computation unit 32 and parallax precision unit 34.In an embodiment, computing module 12 is first in order to calculate the least absolute value sum of the deviations between the first visual angle frame 11 with the second visual angle frame 13 between each corresponding block, and meaning is namely as rear " coupling cost " word used.Absolute value error summation carries out the simplest a kind of art of computation in similarity measurement, its practice is with reference to centered by a certain pixel in image (the first visual angle frame 11) and target image (the second visual angle frame 13), each pixel in the square scope of its periphery next-door neighbour subtracts each other, then get subtract each other result absolute value after carry out cumulative obtaining.
Further, disparity computation unit 32 will select reckling in absolute value error summation.Least absolute value sum of the deviations can be used to look for the block (object) matched.If LOOK LEFT image and LOOK RIGHT image are for conform to completely, then least absolute value sum of the deviations will be 0.Parallax precision unit 34 provides further and makes the mechanism of least absolute value sum of the deviations more precision to produce parallax distribution map, and wherein the method for precision can be reached by various different known method.Therefore, the depth information of image can be derived after the calculating of parallax distribution map completes.
Fig. 4 is in content one embodiment of the present invention, the more detailed calcspar of feature extraction module 14.Feature extraction module 14 receives the block cut out by the first visual angle frame 11 from receiver module 10, and to each onblock executing feature extraction program, to produce multiple characteristic profile, wherein characteristic profile respectively comprises multiple character numerical value, with each block of correspondence.In the present embodiment, feature extraction module 14 comprises chroma and to be as the criterion color segmentation unit 40, edge detection unit 42 and position analysis unit 44.The chroma color segmentation unit 40 that is as the criterion produces color segmentation distribution map, comprises multiple color character numerical value.More specifically, chroma is as the criterion color segmentation unit 40 in order to will the pixel region of analogous color be had to be divided into same group according to chroma information to the pixel distinguished in each block.Group number represents the object number in block.Color character numerical value is namely relevant to the group number in block.
Edge detection unit 42, in order to produce edge distribution figure, comprises multiple edge feature numerical value, the edge pixel quantity wherein in edge feature numeric representation block.Edge pixel can be judged by high pass filter.Wherein, " edge pixel " one word refer to the pixel be positioned on edge.Position analysis unit 44, in order to produce attentiveness distribution map, comprises multiple attentiveness character numerical value.In section Example, setting position analytic unit 44 can not be needed.
Above-mentioned color character numerical value, edge feature numerical value and attentiveness character numerical value represent the characteristic strength of color, edge and attentiveness respectively.In an embodiment, when color character numerical value one of them higher time, the probability in the block of its correspondence with the object different from surrounding environment is namely higher.The reliability of the parallax numerical value of this block is therefore higher.And when edge feature numerical value one of them higher time, the number of edges in the block of its correspondence is then larger.The reliability of the parallax numerical value of this block is also therefore higher.And when attentiveness character numerical value one of them higher time, the position of the block of its correspondence is easier to as observer is viewed.For example, be positioned at the middle block of image owing to being easier to arrive observed by observer, and there is higher attentiveness character numerical value.
It is noted that be only described for the feature of three kinds of forms in above-described embodiment.In other embodiment, also can consider the feature of other form.
Fig. 5 is in content one embodiment of the present invention, the more detailed calcspar of estimating module 16.In the present embodiment, estimating module 16 comprises: reliability calculating unit 50 and comparing unit 52.Wherein reliability calculating unit 50 comprises the similar reliability calculating unit 500 of parallax, edge reliability calculating unit 502, attentiveness reliability calculating unit 504, coupling cost reliability calculating unit 506 and reliability summation unit 508.The coupling cost of aforesaid color segmentation distribution map, edge distribution figure, attentiveness distribution map and block is converted to corresponding reliability distribution map by parallax similar reliability calculating unit 500, edge reliability calculating unit 502, attentiveness reliability calculating unit 504 and coupling cost reliability calculating unit 506 respectively.
The similar reliability calculating unit 500 of parallax to be as the criterion the color segmentation distribution map that color segmentation unit 40 produces and the parallax distribution map that the computing unit 12 of Fig. 1 produces according to the chroma that Fig. 4 illustrates, and produces similar reliability distribution map.The color of a particular block and around it color of block close or in an approximate scope time, the parallax numerical value of the parallax numerical value of this particular block and the block of its periphery compares by the similar reliability calculating unit 500 of parallax, to verify the reliability of parallax numerical value and to carry out normalization to parallax numerical value further.Therefore similar reliability distribution map comprises several similar reliability numerical value, respectively corresponds to the color character numerical value in color segmentation distribution map.Around particular block and its, the color of block is close and also have close parallax numerical value, then the similar reliability numerical value of this particular block is also higher.
The edge distribution figure that the edge detection unit 42 that edge reliability calculating unit 502 illustrates according to Fig. 4 produces produces edge reliability distribution map, wherein reliability distribution map in edge comprises multiple edges reliability numerical value, respectively corresponds to an edge value in edge distribution figure.More specifically, reliability numerical value in edge is proportional to edge value.For example, reliability numerical value in edge can produce by by the numeral of regular for the edge value 0-2 of turning to.
Attentiveness reliability calculating unit 504, in order to the attentiveness distribution map produced according to the position analysis unit 44 illustrated in Fig. 4, produces attentiveness reliability distribution map.Wherein attentiveness reliability distribution map comprises multiple attentiveness reliability numerical value, respectively corresponds to one of them attentiveness intensity of block.
Coupling cost reliability calculating unit 506, according to the coupling cost of each block (that is the disparity computation unit 32 that comprises of the computing module 12 illustrated by Fig. 1 calculate least absolute value sum of the deviations), produces coupling cost reliability distribution map.Wherein mate cost reliability distribution map and comprise multiple coupling cost reliability numerical value, respectively correspond to one of them coupling cost intensity of block.
Therefore, reliability summation unit 508 calculates reliability distribution map by according to above-mentioned characteristic profile.Wherein reliability distribution map comprises the feature reliability of each block.In other words, feature reliability is the sum total of similar reliability numerical value, edge reliability numerical value, coupling cost reliability numerical value and the attentiveness reliability numerical value that block is corresponding.
Comparing unit 52 receives reliability distribution map, to compare according to the feature reliability of each block and at least one reliability critical value, carrys out the defective block of filtering, produces multiple candidate block further.When the feature reliability of a particular block is not high enough, because it is that the probability of recessed block and convex region block is lower, and defective block will be regarded as and give up.
Therefore, comparing unit 52 is judged the parallax numerical value of these candidate block further by parallax distribution map, and according to there is in candidate block maximum disparity numerical value and minimum parallax numerical value person judges recessed block and convex region block respectively.In an embodiment, recessed block corresponds to the maximum disparity numerical value in parallax distribution map, and convex region block corresponds to the minimum parallax numerical value in parallax distribution map.
The Stereo matching device 1 of content of the present invention is the framework according to reliability running, can detect real recessed block and convex region block when not needing high complexity Stereo Matching Technology.By capturing the characteristic information of block, the feature of each block can be calculated rapidly, and it assesses the cost quite low.The reliability of feature is further derived, and uses the candidate block to select to have higher feature reliability.Therefore, recessed block and convex region block can promptly be selected from candidate block again.
Please refer to Fig. 6.Fig. 6 is in content one embodiment of the present invention, a kind of flow chart of solid matching method 600.Solid matching method 600 can be applicable to as Fig. 1 the Stereo matching device 1 that illustrates.Solid matching method 600 comprises the following step (should be appreciated that, step mentioned in the present embodiment, except chatting its order person bright especially, all can adjust its tandem according to actual needs, even can perform simultaneously or partly simultaneously).
In step 601, receiver module 10 receives the first visual angle frame 11 and the second visual angle frame 13.It is multiple block that receiver module 10 splits the first visual angle frame 11 further in step 602.In step 603, computing module 12 produces parallax distribution map, and wherein parallax distribution map comprises multiple parallax numerical value, respectively corresponding block one of them the first visual angle frame 11 and the second visual angle frame 13 between a parallax amount.
In step 604, feature extraction module 14 performs feature extraction program, and to produce multiple characteristic profile, wherein characteristic profile respectively comprises multiple character numerical value, with each block of correspondence.
In step 605, the reliability calculating unit 50 in estimating module 16 calculates the feature reliability of each block according to characteristic profile, and wherein feature reliability is the summation of the character numerical value of each characteristic profile.
In step 606, the comparing unit 52 in estimating module 16 compares with the multiple defective block of filtering according to the feature reliability of each block and at least one reliability critical value, produces multiple candidate block further.Comparing unit 52 further in step 607 according to there is in candidate block maximum disparity numerical value and minimum parallax numerical value person judges recessed block and convex region block respectively.Wherein, the parallax distribution map that the parallax numerical value of candidate block can be produced by step 603 judges.
Although content of the present invention discloses as above with execution mode; so itself and be not used to limit content of the present invention; any those skilled in the art; in the spirit and scope not departing from content of the present invention; when being used for a variety of modifications and variations, therefore the protection range of content of the present invention is when being as the criterion depending on the appended right person of defining.
Claims (14)
1. a Stereo matching device, in a three-dimensional display system to judge a recessed block and a convex region block, wherein this Stereo matching device at least comprises:
One receiver module is multiple block in order to receive one first visual angle frame and one second visual angle frame and to split this first visual angle frame;
One computing module, in order to produce a parallax distribution map, wherein this parallax distribution map comprises multiple parallax numerical value, respectively corresponding the plurality of block one of them this first visual angle frame and a parallax amount of this second visual angle interframe;
One feature extraction module, in order to each the plurality of onblock executing one feature extraction program, to produce multiple characteristic profile, wherein the plurality of characteristic profile respectively comprises multiple character numerical value, with each the plurality of block of correspondence; And
One estimating module, comprises:
One reliability calculating unit, in order to calculate a feature reliability of each the plurality of block according to the plurality of characteristic profile, wherein this feature reliability is a summation of the plurality of character numerical value of each the plurality of characteristic profile; And
One comparing unit, in order to compare with the multiple defective block of filtering according to this feature reliability of each the plurality of block and at least one reliability critical value, the multiple candidate block of further generation, and according to there is in the plurality of candidate block a maximum disparity numerical value and a minimum parallax numerical value person judges this recessed block and this convex region block respectively;
Wherein this recessed block corresponds to this maximum disparity numerical value in this parallax distribution map, and this convex region block corresponds to this minimum parallax numerical value in this parallax distribution map.
2. Stereo matching device according to claim 1, wherein this feature extraction module comprises:
One chroma is as the criterion color segmentation unit, produces a color segmentation distribution map, comprise multiple color character numerical value with one; And
One edge detection unit, in order to produce fate Butut, comprises multiple edge feature numerical value.
3. Stereo matching device according to claim 2, wherein this feature extraction module also comprises a position analysis unit, in order to produce an attentiveness distribution map, comprises multiple attentiveness character numerical value.
4. Stereo matching device according to claim 3, wherein this reliability calculating unit comprises:
The similar reliability calculating unit of one parallax, in order to produce a similar reliability distribution map according to this color segmentation distribution map and this parallax distribution map, wherein this similar reliability distribution map comprises multiple similar reliability numerical value, respectively correspond to the plurality of block one of them;
One edge reliability calculating unit, in order to produce an edge reliability distribution map according to this edge distribution figure, wherein this edge reliability distribution map comprises multiple edges reliability numerical value, respectively correspond to the plurality of block one of them; And
Coupling cost reliability calculating unit, in order to produce a coupling cost reliability distribution map according to a least absolute value sum of the deviations of each the plurality of block, wherein this coupling cost reliability distribution map comprises multiple coupling cost reliability numerical value, respectively corresponds to one of them a coupling cost intensity of the plurality of block;
Wherein this feature reliability is the sum total of the plurality of similar reliability numerical value, the plurality of edge reliability numerical value and the plurality of coupling cost reliability numerical value that the plurality of block is corresponding.
5. Stereo matching device according to claim 4, wherein this reliability calculating unit comprises an attentiveness reliability calculating unit, in order to produce an attentiveness reliability distribution map according to this attentiveness distribution map, wherein this attentiveness reliability distribution map comprises multiple attentiveness reliability numerical value, respectively correspond to one of them an attentiveness intensity of the plurality of block, this feature reliability is the plurality of similar reliability numerical value that the plurality of block is corresponding, the plurality of edge reliability numerical value, the sum total of the plurality of coupling cost reliability numerical value and the plurality of attentiveness reliability numerical value.
6. Stereo matching device according to claim 1, wherein this computing module calculates a least absolute value sum of the deviations to produce this parallax distribution map according to each multiple corresponding blocks in each the plurality of block of this first visual angle frame and this second visual angle frame.
7. Stereo matching device according to claim 1, wherein this receiver module also comprises a low pass filter and a minimizing sampler, reduces sample program to perform a low-pass filtering program and to this first visual angle frame and this second visual angle frame respectively.
8. a solid matching method, for the Stereo matching device in a three-dimensional display system, to judge a recessed block and a convex region block, wherein this solid matching method at least comprises:
Receive one first visual angle frame and one second visual angle frame;
Splitting this first visual angle frame is multiple block;
Produce a parallax distribution map, wherein this parallax distribution map comprises multiple parallax numerical value, respectively corresponding the plurality of block one of them this first visual angle frame and a parallax amount of this second visual angle interframe;
Perform a feature extraction program, to produce multiple characteristic profile, wherein the plurality of characteristic profile respectively comprises multiple character numerical value, with each the plurality of block of correspondence;
Calculate a feature reliability of each the plurality of block according to the plurality of characteristic profile, wherein this feature reliability is a summation of the plurality of character numerical value of each the plurality of characteristic profile;
Compare with the multiple defective block of filtering according to this feature reliability of each the plurality of block and at least one reliability critical value, produce multiple candidate block further; And
According to there is in the plurality of candidate block a maximum disparity numerical value and a minimum parallax numerical value person judges this recessed block and this convex region block respectively;
Wherein this recessed block corresponds to this maximum disparity numerical value in this parallax distribution map, and this convex region block corresponds to this minimum parallax numerical value in this parallax distribution map.
9. solid matching method according to claim 8, the step wherein performing this feature extraction program also comprises:
Produce a color segmentation distribution map, comprise multiple color character numerical value; And
Produce fate Butut on one side, comprise multiple edge feature numerical value.
10. solid matching method according to claim 9, the step wherein performing this feature extraction program also comprises:
Produce an attentiveness distribution map, comprise multiple attentiveness character numerical value.
11. solid matching methods according to claim 10, the step wherein calculating this feature reliability also comprises:
Produce a similar reliability distribution map according to this color segmentation distribution map and this parallax distribution map, wherein this similar reliability distribution map comprises multiple similar reliability numerical value, respectively correspond to the plurality of block one of them;
Produce an edge reliability distribution map according to this edge distribution figure, wherein this edge reliability distribution map comprises multiple edges reliability numerical value, respectively correspond to the plurality of block one of them;
A coupling cost reliability distribution map is produced according to a least absolute value sum of the deviations of each the plurality of block, wherein this coupling cost reliability distribution map comprises multiple coupling cost reliability numerical value, respectively corresponds to one of them a coupling cost intensity of the plurality of block; And
The plurality of similar reliability numerical value, the plurality of edge reliability numerical value and the plurality of coupling cost reliability numerical value corresponding to the plurality of block add up to produce this feature reliability.
12. solid matching methods according to claim 11, the step wherein calculating this feature reliability also comprises:
An attentiveness reliability distribution map is produced according to this attentiveness distribution map, wherein this attentiveness reliability distribution map comprises multiple attentiveness reliability numerical value, respectively correspond to one of them an attentiveness intensity of the plurality of block, this feature reliability is the sum total of the plurality of similar reliability numerical value that the plurality of block is corresponding, the plurality of edge reliability numerical value, the plurality of coupling cost reliability numerical value and the plurality of attentiveness reliability numerical value.
13. solid matching methods according to claim 8, the step wherein producing this parallax distribution map also comprises: calculate a least absolute value sum of the deviations according to each multiple corresponding blocks in each the plurality of block of this first visual angle frame and this second visual angle frame.
14. solid matching methods according to claim 8, the step wherein receiving one first visual angle frame and one second visual angle frame also comprises and performs a low-pass filtering program and to this first visual angle frame and this second visual angle frame respectively and reduce sample program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210077582.4A CN103327346B (en) | 2012-03-22 | 2012-03-22 | Stereo matching device for judging concave and protruding blocks and method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210077582.4A CN103327346B (en) | 2012-03-22 | 2012-03-22 | Stereo matching device for judging concave and protruding blocks and method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103327346A CN103327346A (en) | 2013-09-25 |
| CN103327346B true CN103327346B (en) | 2015-04-01 |
Family
ID=49195840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210077582.4A Active CN103327346B (en) | 2012-03-22 | 2012-03-22 | Stereo matching device for judging concave and protruding blocks and method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103327346B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI637350B (en) | 2018-01-09 | 2018-10-01 | 緯創資通股份有限公司 | Method, image processing device, and system for generating disparity map |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7227526B2 (en) * | 2000-07-24 | 2007-06-05 | Gesturetek, Inc. | Video-based image control system |
| CN102074014A (en) * | 2011-02-23 | 2011-05-25 | 山东大学 | Stereo matching method by utilizing graph theory-based image segmentation algorithm |
| CN102223556A (en) * | 2011-06-13 | 2011-10-19 | 天津大学 | Multi-view stereoscopic image parallax free correction method |
| CN102263957A (en) * | 2011-07-25 | 2011-11-30 | 北京航空航天大学 | Search-window adaptive parallax estimation method |
-
2012
- 2012-03-22 CN CN201210077582.4A patent/CN103327346B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7227526B2 (en) * | 2000-07-24 | 2007-06-05 | Gesturetek, Inc. | Video-based image control system |
| CN102074014A (en) * | 2011-02-23 | 2011-05-25 | 山东大学 | Stereo matching method by utilizing graph theory-based image segmentation algorithm |
| CN102223556A (en) * | 2011-06-13 | 2011-10-19 | 天津大学 | Multi-view stereoscopic image parallax free correction method |
| CN102263957A (en) * | 2011-07-25 | 2011-11-30 | 北京航空航天大学 | Search-window adaptive parallax estimation method |
Non-Patent Citations (1)
| Title |
|---|
| A Stereoscopic Video Generation Method Using Stereoscopic Display Characterization and Motion Analysis;Dongbo Min eta.;《IEEE TRANSACTIONS ON BROADCASTING》;20080630;第54卷(第2期);188-197 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103327346A (en) | 2013-09-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110544258B (en) | Image segmentation method and device, electronic equipment and storage medium | |
| CN111242097B (en) | Face recognition method and device, computer readable medium and electronic equipment | |
| CN109076198B (en) | Video-based object tracking occlusion detection system, method and equipment | |
| US8989481B2 (en) | Stereo matching device and method for determining concave block and convex block | |
| JP6496987B2 (en) | Target detection method and target detection apparatus | |
| Rossi et al. | On the use of stereovision to develop a novel instrumentation system to extract geometric fire fronts characteristics | |
| US8588515B2 (en) | Method and apparatus for improving quality of depth image | |
| EP2860695B1 (en) | Method and apparatus for identifying noise in disparity depth image | |
| CN101443817B (en) | Method and device for determining correspondence, preferably for the three-dimensional reconstruction of a scene | |
| CN113221677B (en) | Track abnormality detection method and device, road side equipment and cloud control platform | |
| CN106373162B (en) | A kind of obvious object detection method merged based on conspicuousness with propagation | |
| EP3073443A1 (en) | 3D Saliency map | |
| CN105469380A (en) | Method and device for detecting shielding against object | |
| CN105740751A (en) | Object detection and identification method and system | |
| CN114565908A (en) | Lane line detection method and device, electronic device and storage medium | |
| CN107392948B (en) | Image registration method of amplitude-division real-time polarization imaging system | |
| CN103327346B (en) | Stereo matching device for judging concave and protruding blocks and method thereof | |
| CN105138979A (en) | Method for detecting the head of moving human body based on stereo visual sense | |
| Srikakulapu et al. | Depth estimation from single image using defocus and texture cues | |
| Rossi et al. | Dynamic fire 3D modeling using a real-time stereovision system | |
| Kim et al. | A novel red apple detection algorithm based on adaboost learning | |
| KR101501531B1 (en) | Stereo Vision-based Pedestrian Detection System and the method of | |
| KR101357581B1 (en) | A Method of Detecting Human Skin Region Utilizing Depth Information | |
| CN110910379B (en) | Incomplete detection method and device | |
| CN114549584A (en) | Information processing method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |